Mechanical degradations of Fe–C alloys induced by stress corrosion in supercritical CO₂ environments: a study based on molecular dynamics simulation and machine learning

Under supercritical carbon dioxide (SCO₂) conditions, stress corrosion cracking (SCC) of steel can significantly degrade their mechanical properties and shorten their service life. However, few studies have been focused on predicting such property degradation. This study investigates the degradation of mechanical properties of Fe–C alloys induced by SCC in SCO₂ environments by using ReaxFF molecular dynamics (MD) simulation and machine learning (ML) algorithms. The considerations are given to the effect of factors including C content in the Fe–C alloy, the density of CO₂, temperature, and strain rate. These factors are varied to generate 625 different conditions. MD results reveal that the corrosion happens with O atoms penetrating deeply into the metal while C atoms aggregated in the subsurface layer to form a carbon/iron mixed region, inhibiting further corrosion. However, the applied tensile stress exposes the new metal matrix to the CO₂ environment and thereby facilitates further corrosion reactions. It is interesting that initial stress in the tensile direction is negative at the beginning of the tensile deformation, which is due to that the corrosion induces compressive stress in the corroded regions. Based on the stress–strain curves obtained in MD simulation, key mechanical parameters are extracted, including the initial stress before tensile deformations, modulus of elasticity, and ultimate tensile strength. The best-performing ML models for each parameter are trained and demonstrate good predictive performance in unknown conditions with R² > 0.9. The theoretical framework proposed can assist in evaluating the performance and service life of materials in extreme environments.